Crusoe Energy
Last reviewed
Jun 7, 2026
Sources
No citations yet
Review status
Needs citations
Revision
v2 ยท 4,060 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 7, 2026
Sources
No citations yet
Review status
Needs citations
Revision
v2 ยท 4,060 words
Add missing citations, update stale details, or suggest a clearer explanation.
Crusoe Energy Systems, Inc. (commonly branded as Crusoe) is a privately held American energy and computing infrastructure company headquartered in Denver, Colorado. Founded in 2018 by Chase Lochmiller and Cully Cavness, the company originally combined oilfield natural gas, modular containerized data centers, and cryptocurrency mining to monetize "stranded" gas that would otherwise be flared at remote well sites. Beginning in 2023, Crusoe pivoted toward generative AI and high-performance computing, building data center capacity for large GPU clusters under the brand "Crusoe Cloud" and constructing the flagship Abilene, Texas campus that anchors the Stargate Project joint venture with OpenAI, Oracle, and SoftBank.[1][2][3] In March 2025 Crusoe divested its bitcoin mining business to NYDIG to focus exclusively on AI infrastructure.[4][5]
| Item | Detail |
|---|---|
| Legal name | Crusoe Energy Systems, Inc. |
| Founded | 2018, Denver, Colorado |
| Founders | Chase Lochmiller (CEO), Cully Cavness (President) |
| Headquarters | Denver, Colorado, United States |
| Industries | AI cloud, data center development, energy infrastructure |
| Notable products | Crusoe Cloud, Digital Flare Mitigation (DFM) modules (divested 2025) |
| Series E valuation | Above $10 billion (October 2025) |
| Lead Series D investor | Founders Fund (December 2024) |
| Lead Series E investors | Valor Equity Partners, Mubadala Capital (October 2025) |
Crusoe was founded in 2018 in Denver, Colorado. The two co-founders had complementary backgrounds. Chase Lochmiller had studied mathematics and physics at MIT and computer science at Stanford, traded at quantitative firms including Jump Trading and GETCO, and then served as a general partner at Polychain Capital, a crypto-focused investment firm.[6][7] Cully Cavness had worked in oil and gas investment banking at Petrie Partners and at Highlands Natural Resources after earning a geology degree from Middlebury College and an MBA from the University of Oxford.[6][7] The two had known each other since high school in Colorado.[6]
The original business idea united their experiences. Oil wells in shale basins routinely produce associated natural gas as a byproduct. Where pipeline takeaway capacity is missing, that gas is typically flared, that is, burned at the wellhead, releasing carbon dioxide and slip methane into the atmosphere. Crusoe proposed to install gas-fired reciprocating generators next to the wellhead and use the electricity to power modular shipping-container data centers, which it called Digital Flare Mitigation (DFM) systems. The data centers ran compute workloads including cryptocurrency mining, monetizing energy that would otherwise be wasted.[7][8]
The company raised a $4.5 million seed round in March 2019 and a $30 million Series A in December 2019 led by Bain Capital Ventures.[9]
Through 2020 and 2021 Crusoe scaled rapidly across the Bakken, Powder River, Permian, and other oil-producing basins. In April 2021 the company closed a $128 million Series B led by Valor Equity Partners, with participation from Lowercarbon Capital, DRW Venture Capital, Founders Fund, Bain Capital Ventures, Coinbase Ventures, Polychain Capital, Upper90, Winklevoss Capital, Exor, Zigg Capital, and J.B. Straubel, among others.[10] Crusoe used Waukesha (INNIO) reciprocating gas engines packaged with its modular data centers to convert raw associated gas into electricity.[11] Crusoe stated that its stoichiometric combustion design achieved approximately 99.9% combustion efficiency, compared with an industry average of about 91.1% for open flares, reducing methane slip by roughly 99% and CO2-equivalent emissions by up to about 68.6% relative to flaring.[11]
In April 2022 Crusoe announced a $350 million Series C equity financing led by climate-focused investor G2 Venture Partners, plus credit facilities expandable to $155 million from SVB Capital, Sparkfund, and Generate Capital, bringing total new capital to $505 million. The company said it would use the proceeds to expand its DFM fleet and to launch Crusoe Cloud, a low-cost high-performance computing platform.[12][13]
In June 2022 Crusoe acquired Easter-Owens Electric Co., a Denver-area modular data center and electrical systems manufacturer that had been its primary supplier since 2020. The acquisition vertically integrated data center design, fabrication, and switchgear manufacturing under one roof; roughly 70 Easter-Owens employees joined Crusoe's then-172 person team.[14]
By 2023 the economics of large-model AI training shifted Crusoe's strategic focus. In October 2023 Crusoe announced that it had selected HPE Cray XD supercomputers populated with NVIDIA H100 Tensor Core GPUs and NVIDIA Quantum-2 InfiniBand interconnect to build out Crusoe Cloud for generative-AI training and tuning. The same announcement described plans to deploy NVIDIA L40S GPUs for generative-AI inference, graphics, and scientific computing.[15][16] In a parallel announcement Crusoe disclosed a $200 million GPU purchase commitment from Upper90 to fund continued cloud expansion.[17]
On July 18, 2024, Crusoe announced that it would build an initial 200 MW AI data center at the Lancium Clean Campus outside Abilene, Texas, with plans to expand to a total 1.2 GW of capacity in subsequent phases. Crusoe described each building as engineered to support up to 100,000 GPUs on a single integrated network fabric, with direct-to-chip liquid cooling and provisions for rear-door heat exchangers and air cooling.[18]
On October 15, 2024, Crusoe, Blue Owl Capital, and Primary Digital Infrastructure announced a $3.4 billion joint venture to fund the first two buildings of the Abilene campus, a 206 MW two-building facility totaling roughly 998,000 square feet, with Crusoe acting as designer, builder, and operator and Blue Owl and Primary Digital Infrastructure providing forward-takeout capital.[19][20]
On December 12, 2024, Crusoe announced the closing of a $600 million Series D equity round led by Founders Fund at a $2.8 billion post-money valuation. Participating investors included Fidelity, Long Journey Ventures, Mubadala, NVIDIA, Ribbit Capital, and Valor Equity Partners.[1][21] The company said proceeds would fund expansion across its value chain, including data centers and Crusoe Cloud.[1]
On January 21, 2025, U.S. President Donald Trump appeared at a White House press conference alongside Sam Altman of OpenAI, Larry Ellison of Oracle, and Masayoshi Son of SoftBank to announce the Stargate Project, a planned $500 billion AI infrastructure joint venture in the United States, with an initial $100 billion committed up front. Stargate LLC was structured as a joint venture among OpenAI, SoftBank, Oracle, and Abu Dhabi-based MGX, with SoftBank and OpenAI as the financial and operational leads respectively. Initial technology partners listed at the announcement were Arm, Microsoft, NVIDIA, Oracle, and OpenAI.[22][23][24]
The Abilene campus that Crusoe had begun building in mid-2024 became the flagship Stargate Site 1. Crusoe owns and operates the buildings and leases them to Oracle, which in turn provides Oracle Cloud Infrastructure (OCI) capacity to OpenAI for training and inference. Larry Ellison stated at the press event that ten buildings were being built in Abilene with plans to expand.[22][23]
On March 25, 2025, Crusoe announced an agreement to divest its bitcoin mining business to NYDIG, the bitcoin subsidiary of Stone Ridge Holdings. The deal covered approximately 270 MW of power generation and more than 425 modular data centers across the United States and Argentina, along with about 135 employees who transferred to NYDIG. Financial terms were not disclosed, but Crusoe was reported to become a major equity holder in the combined NYDIG entity, second only to Stone Ridge. The transaction allowed Crusoe to focus exclusively on AI infrastructure.[4][5][25]
On March 18, 2025, Crusoe announced that construction had begun on a second phase at the Abilene campus, six additional buildings that, when complete in mid-2026, would bring the campus to eight buildings, approximately 4 million square feet, and 1.2 GW of total power capacity.[26] In May 2025 the Blue Owl and Primary Digital Infrastructure joint venture was expanded with additional financing, including roughly $11.6 billion arranged via JPMorgan Chase and other capital, to fund the campus build-out.[27]
On October 24, 2025, Crusoe announced the initial closing of an oversubscribed $1.375 billion Series E co-led by Valor Equity Partners and Mubadala Capital at a valuation above $10 billion. Participants included NVIDIA, Founders Fund, Fidelity Management & Research Company, T. Rowe Price, and Tiger Global Management. The company disclosed that Crusoe Cloud bookings had grown five-fold year over year through the first three quarters of 2025 and that it was building multiple gigawatt-scale campuses, including a 1.8 GW site in Wyoming.[28][29]
DFM was Crusoe's original product family and the basis for the "Energy" in its corporate name. A DFM module is a containerized unit that takes raw associated gas, processes and dries it, feeds it to one or more reciprocating gas engines (the company has publicly named INNIO Waukesha gensets as a key supplier), and uses the electrical output to power a co-located modular data center on the same well pad. Module ratings published by Crusoe ranged from systems sized for roughly 70,000 standard cubic feet per day (scfd) of gas up to about 300,000 scfd individually, deployable in series for higher volumes.[11][30]
The technical case for DFM rested on three quantitative claims. First, conventional open flares are imperfect combustors, particularly when wind speeds, fuel composition, or operating conditions move outside the optimum window; field measurement studies cited by Crusoe and others put average flare combustion efficiency at approximately 91.1%, meaning roughly 9% of fuel methane escapes uncombusted. Second, methane is a much more potent greenhouse gas than CO2 over short time horizons. Third, a stoichiometric, controlled engine combustion environment can achieve combustion efficiency near 99.9%, materially reducing methane slip and shrinking the warming impact even though the resulting CO2 is itself counted as a greenhouse gas.[11][30][38]
By the time of the 2022 Series C announcement Crusoe reported a deployed fleet capable of avoiding emissions equivalent to roughly 650,000 metric tons of CO2 per year, comparable to taking about 140,000 cars off the road.[12] In later public materials Crusoe described running approximately 125 modular data centers across U.S. oilfields and processing approximately 20 million cubic feet of gas per day, claiming approximately 800,000 tons per year of CO2-equivalent reductions.[30] The entire DFM business and the related modular data centers, totaling about 270 MW and 425 units, were transferred to NYDIG in 2025 along with roughly 135 employees, leaving Crusoe entirely focused on AI infrastructure thereafter.[4][5]
Crusoe Cloud is the company's GPU Infrastructure-as-a-Service platform. The cloud catalog at the time of the 2023 launch was built on HPE Cray XD supercomputers configured with NVIDIA H100 GPUs and Quantum-2 InfiniBand fabric; the catalog later expanded to NVIDIA H200 HGX instances aimed at large-context generative AI training and inference. Crusoe also publicly announced support for AMD Instinct MI300X and later-generation accelerators, alongside NVIDIA Blackwell GB200 NVL72 systems delivered into the Abilene campus.[15][16][31][32]
Architecturally, Crusoe Cloud is differentiated from generic infrastructure-as-a-service offerings by its tight coupling with Crusoe-owned data centers and, in many sites, Crusoe-owned generation. The platform is oriented around large reserved capacity contracts rather than per-minute consumption, reflecting the workload mix (training large foundation models, fine-tuning, and high-throughput batch inference) that typically requires thousands of GPUs co-located on a single high-bandwidth fabric. Crusoe's network design publicly emphasizes large non-blocking InfiniBand fabrics within each building, supporting tightly synchronized collective operations such as all-reduce that dominate large-model training. By the time of the Series E disclosure in October 2025, the company reported that Crusoe Cloud bookings had grown roughly five-fold year over year in the first three quarters of 2025, an indicator of the demand absorption that the new Abilene capacity was designed to serve.[28][29]
The Abilene campus is Crusoe's flagship project and the physical anchor of the Stargate Project. As publicly disclosed, the first two buildings span roughly 998,000 square feet and approximately 206 MW of IT capacity, with the full eight-building campus targeting roughly 4 million square feet and 1.2 GW of capacity at completion in mid-2026. Site preparation began in June 2024; the first NVIDIA GB200 racks were delivered in mid-2025; and Crusoe announced that the first phase was live on Oracle Cloud Infrastructure on September 30, 2025, supporting early training and inference workloads.[18][20][26][33]
The Abilene campus is situated on the Lancium Clean Campus, a 1,100-acre site that Lancium had been developing in West Texas since 2020. The location was selected for proximity to West Texas renewable generation and for available transmission interconnection. Each building is engineered for direct-to-chip liquid cooling at GPU-rack densities far above conventional enterprise air-cooled facilities, with each building publicly described by Crusoe as capable of operating up to 100,000 GPUs on a single integrated network fabric. The structural and electrical design allows individual buildings to be brought online incrementally as Oracle delivers successive racks of NVIDIA accelerators, a sequencing pattern intended to match construction with the cadence at which hyperscale buyers can absorb new capacity for model training.[18][20][33]
In addition to its original flare-gas approach, Crusoe has publicly emphasized siting data centers near "abundant" or "stranded" energy sources, including dedicated natural gas generation, wind, and geothermal. The company's European footprint includes capacity at atNorth's ICE02 data center in Iceland, powered by geothermal and hydroelectric energy, and an announced site in Norway through a partnership with Polar.[34][35] In 2025 Crusoe also disclosed plans for behind-the-meter natural gas generation at multi-gigawatt scale to fuel new AI campuses, while continuing to position renewable-aligned sites in its public messaging.[36]
The energy strategy is intertwined with the company's identity as an "AI factory" developer rather than a tenant. By owning the power generation, the modular building, and the white-space buildout, Crusoe argues that it can both compress the schedule between site identification and racked GPUs and capture economic rents that conventional colocation tenants leave to landlords and utilities. The same vertical integration introduces concentration risk: a large share of revenue and forward capacity is tied to a small number of hyperscale anchor tenants, principally Oracle (for the Stargate-aligned capacity at Abilene) and, in subsequent expansions, additional hyperscale buyers.[3][33][36]
| Round | Date | Amount | Lead investor(s) | Valuation |
|---|---|---|---|---|
| Seed | March 2019 | $4.5M | Founders Fund (among others) | not disclosed |
| Series A | December 2019 | $30M | Bain Capital Ventures | not disclosed |
| Series B | April 2021 | $128M | Valor Equity Partners | not disclosed |
| Series C | April 2022 | $350M equity (+$155M credit) | G2 Venture Partners | $1.75B (per secondary reporting) |
| Series D | December 12, 2024 | $600M | Founders Fund | $2.8B |
| Series E | October 24, 2025 | $1.375B | Valor Equity Partners and Mubadala Capital | above $10B |
Sources for the table: Crusoe announcements and secondary press coverage cited at [1][9][10][12][28][29].
Chase Lochmiller has served as chief executive officer and as a co-founder of Crusoe since 2018. He previously was a general partner at Polychain Capital and, earlier, a quantitative trader at Jump Trading and GETCO. He holds undergraduate degrees in mathematics and physics from MIT and a master's in computer science from Stanford with a specialization in artificial intelligence.[6][7] Cully Cavness has served as co-founder, president, and (variously) chief operating officer and chief strategy officer; he previously worked in oil and gas investment banking and at energy producers, holds a geology bachelor's from Middlebury and an MBA from Oxford, and was a Thomas J. Watson Fellow studying Iceland's geothermal industry.[6][7]
Crusoe has consistently presented itself as an "energy-first" or "vertically integrated" AI infrastructure company, contrasting itself with bare-metal GPU rental providers such as CoreWeave and Lambda Labs by emphasizing its ownership of underlying power generation, modular site design, and data-center construction. Its early positioning was specifically environmental: by burning otherwise vented or flared methane in engine-driven combustion, the company argued that it converted a potent greenhouse gas (methane is roughly 80 times more potent than CO2 on a 20-year basis) into the less-potent CO2 produced by complete combustion, while simultaneously monetizing the energy.[7][11][37]
As the company shifted from bitcoin to AI workloads, its differentiation narrative also shifted toward speed of build, ability to site capacity in power-rich but transmission-constrained areas, and vertical control over the building shell, electrical balance-of-plant, and white-space fit-out. Public reporting on the Abilene campus emphasized that this control, combined with Crusoe's earlier experience scaling modular sites in the oilpatch, was a meaningful contributor to the speed at which the Stargate flagship was built out.[3][18][33]
The company also frames its energy posture as a procurement advantage: by securing dedicated generation and long-lead grid interconnection in advance, Crusoe argues that it can offer hyperscale buyers a more credible delivery date than competitors that rely on traditional utility interconnect queues. That posture is not unique in the sector, but the combination of energy-first sourcing, vertically integrated manufacturing (legacy Easter-Owens), and large balance-sheet partnerships through Blue Owl and Primary Digital Infrastructure positions Crusoe as a hybrid of a power developer, a data-center builder, and a specialized GPU cloud.[14][19][20]
Crusoe's flare-gas model attracted skepticism from environmental advocates throughout 2021 to 2024. Critics argued that providing oil and gas producers with a profitable use for otherwise stranded gas could indirectly extend the economic life of fossil-fuel extraction; that combustion in field engines is not perfectly clean, so some methane still escapes; and that crypto mining in particular was a controversial end-use for the resulting electricity. Counterarguments from Crusoe and supportive investors emphasized that the alternative was not "no extraction" but flaring with materially higher methane slip, and that combustion efficiency in engines is empirically higher than in open flares.[37][38]
The shift to AI compute reframed some of these debates but did not eliminate them. Crusoe's 2025 disclosures included multi-gigawatt natural-gas generation behind the meter for AI campuses, prompting renewed coverage of the carbon intensity of large model training and inference, even as the company continued to expand renewable-sourced sites such as Iceland and to pursue geothermal-aligned development.[34][36]
| Provider | Core business | Distinguishing feature |
|---|---|---|
| Crusoe | Vertically integrated AI cloud and data-center developer | Owns site design, build, and (historically) on-site power generation; flagship Stargate Abilene campus |
| CoreWeave | Specialty GPU cloud at hyperscale | Large NVIDIA H100, H200, and Blackwell fleet; deep contracts with model labs |
| Lambda Labs | GPU cloud and on-prem appliances | Long-standing AI-focused cloud; on-demand and reserved GPU offerings |